Brendan Gregg

Brendan created flame graphs to make complex performance measurements understandable for engineers investigating bottlenecks. Reusable tools and analysis methods help teams across organisations locate costly execution paths and focus improvement work. Extensions for graphics processors, alongside public writing about performance engineering, broaden the contribution to contemporary workloads and shared engineering practice.

Brendan Gregg is a performance engineer at OpenAI, joining in February 2026 to focus initially on ChatGPT performance. Having worked across operating systems and large cloud services, Brendan develops tools and methods for understanding where software spends time and resources, tackling the growing cost of AI datacentres and exploring engineering methods to find larger performance gains more quickly.

As the creator of flame graphs—visualisations that help engineers identify expensive paths through running software—Brendan makes complex performance data easier to interpret. These tools support the practical work of finding bottlenecks and deciding where to direct improvements. Reusable analysis methods influence performance work across organisations, while GPU extensions bring flame graphs to workloads running on graphics processors. Public writing about performance tools, methods and engineering challenges shares this experience widely, helping teams diagnose performance issues and use computing resources more efficiently.